No taxid when query assigned to a specific seqID (entry is in the taxonomy)

[Piplopp]

Hello !

I'm trying to index the SILVA database for centrifuge. I built the acc2taxid map, the nodes.dmp and names.dmp files just fine but I noticed some sequences were assigned a taxid of 0 when trying to classify. I had the same behavior when using those files but produced by Kraken2.

readID	seqID	taxID	score	2ndBestScore	hitLength	queryLength	numMatches
6d3e5d85b2aa35d58347bb4b9b203e43	U92195.1.1541	0	60025	60025	260	260	6
6d3e5d85b2aa35d58347bb4b9b203e43	MF457876.1.1456	0	60025	60025	260	260	6
6d3e5d85b2aa35d58347bb4b9b203e43	genus	46463	60025	60025	260	260	6
6d3e5d85b2aa35d58347bb4b9b203e43	CS600365.8.1528	0	60025	60025	260	260	6
6d3e5d85b2aa35d58347bb4b9b203e43	U88546.1.1541	0	60025	60025	260	260	6
6d3e5d85b2aa35d58347bb4b9b203e43	EU775002.1.1308	0	60025	60025	260	260	6
-	-	-	-	-	-	-	-
cb6ee53401962a26788af21de2a16f67	AP017610.4186840.4188378	46463	169744	0	427	427	1

As you can see, for the readID: 6d3e5d85b2aa35d58347bb4b9b203e43 a lot of the matches have a seqID assigned but the taxID is 0. I double checked both the centrifuge-build command (no missing taxonomy id in the output) and the various files and all seems to be fine.

The query readID: cb6ee53401962a26788af21de2a16f67 at the end is behaving as expected, the seqID does have it's expected taxid.

For instance for the seqID U92195.1.1541 and MF457876.1.1456

acc2taxid:

U92195.1.1541   46474
MF457876.1.1456 46465

nodes.dmp

46474   |   46454   |   genus   |   -   |
46465   |   46454   |   genus   |   -   |

names.dmp

46474   |   Salmonella  |   -   |   scientific name |
46465   |   Klebsiella  |   -   |   scientific name |

And from centrifuge-inspect:

>U88546.1.1541 Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacterales;Enterobacteriaceae;Salmonella;Salmonella enterica subsp. enterica serovar Paratyphi A
>MF457876.1.1456 Bacteria;Proteobacteria;Gammaproteobacteria;Enterobacterales;Enterobacteriaceae;Klebsiella;Escherichia coli

I also tried to replace the dots in the sequence ids by '_' just in case but the result is the same. I would expect a taxid of 0 if the sequence was not found in the taxonomy or if the seqID was actually a LCA like the third assignment to 'genus', but in those case, my query was assigned to a specific seqID and thus I would expect to find the related taxID.

Maybe there's something I don't fully understand, but in any case if you have any idea of what can be happening or why :)

Thanks a lot !

[cjalder]

Have you had any luck with this? I have a similar problem, however mine does report errors of missing taxonomy ID, even though they exist in the acc2id.map, nodes.dmp and names.dmp files

[sanderdebacker]

Same problem here, with a custom built database. SeqID gets assigned, but without taxID, even though its present in the .map and .dmp files. Haven't found a solution or explanation yet.

[Piplopp]

I noticed that if you set the -k option high enough the taxids appears even for the SeqIDs that were previously problematic (in my case I tried with -k 1000 but I did not manage to understand what's happening so far

[cjalder]

When inspecting my .map file, I noticed a number of formatting issues within it, such as line breaks and merges within a taxID/seqID (A quick way to check for me was too see if any lines started with a number). I was able to resolve my issue by creating a .map file manually. Hope it helps anyone out there!

[sanderdebacker]

Did some steps again and everything seems the same, but the problem has been resolved.

• merge all fasta sequences again
• make my custom .map file again
• build centrifuge index again

Either one of them, or a combination caused the problem to be resolved.

[Piplopp]

Did the exact same thing and the problem has been resolved. What happened is still a mystery